KubeVirt and the VM-Container Convergence

KubeVirt v1.9.0 (July 30, 2026) advances the VM-container convergence. The CNCF Incubating project runs virtual machines natively on Kubernetes alongside containers — and flips a meaningful switch: all Beta feature gates are on by default (VEP 229), with opt-out via disabledFeatureGates.

The pitch is concrete: keep existing VM workloads as VirtualMachine custom resources while containerizing the rest on your own schedule. No separate toolchain, no big-bang migration. v1.9 adds cross-architecture VM execution (Alpha — ARM64 guest on AMD64 via QEMU TCG), zstd-compressed live migration, storage-agnostic incremental backups via Changed Block Tracking, confidential computing (Intel TDX, AMD SEV Beta, IBM Secure Execution GA), and AI/GPU acceleration with NUMA-aware placement and vGPU live migration.

Enterprise proof: NTT Docomo united eight data centers across 2,000 km into one GPU cluster on KubeVirt; Swisscom runs a sovereign cloud on KubeVirt with Kube-OVN. CNCF graduation is now clearly in view.

One control plane for VMs and pods, gradual migration not a cliff. Operator caveat: carefully review the feature-gate report JSON before upgrading to v1.9 and pin disabledFeatureGates to avoid any surprises.